Sleep changes in late-life depression may be permanent or may recover only slowly. Hence, to clarify the relation of persistent sleep abnormalities, pathogenesis, and illness course, we propose three related longitudinal studies of: Study I - 200 elderly(60-80 yo) patients with recurrent unipolar depression participating in an already funded maintenance therapies research protocol (MH43832-01, C.F. Reynolds, .P. I.) ; Study 2 - 80 elderly subjects (60-80 yo) with spousal bereavement (n=40) and bereavement-related depression (n=40); and Study 3 - 60 healthy elderly (60-80 yo) non-depressed, non-bereaved controls. Participants in all three studies will have repeated (four to seven) sleep EEG assessments over a two to three year period to determine if, as hypothesized: la)abnormal pretreatment (T1) measures of REM sleep cyclicity/density predict elevated risk of relapse and recurrence of major depression; lb)EEG sleep abnormalities persisting into continuation therapy (T2) and into maintenance therapy (T3-T7) predict elevated risk of relapse and recurrence, respectively; 2a)depressed bereaved differ from non-depressed bereaved elderly in baseline (T1) and followup (T2-T5) measures of REM sleep cyclicity/density; 2b)baseline (T1) and followup (T2-T5) abnormalities of REM sleep in bereavement predict a more protracted course (and need for psychiatric intervention); and 3)longitudinal changes in sleep/sleep quality (T1-T4) of healthy elderly controls are related to subclinical depressed mood. Multiple logistic regression and survival analyses will model the hypothesized relation of REM sleep abnormalities (independent variable) to outcome: recurrence/ non-recurrence of depression (Study 1); or presence/absence of RDC-defined major depression ,at 1 and at 2 years after spousal bereavement. A repeated-measures MANOVA will contrast the long-term evolution of selected sleep variables (sensitive to aging, depressed mood, or both) in recurrent depression, bereavement, bereavement-related depression and controls. Exploratory data analyses will assess possible interrelations among biological REM sleep abnormalities, severity of psychopathology, and disruption in social rhythms.